AI Tool Automates Variant Curation and Interpretation for Cancer Whole-Genome Sequencing

Cancer whole-genome sequencing (WGS) provides a comprehensive view of cancer genomes, but interpreting the vast amount of data remains challenging. As WGS gains wider clinical adoption, variant curation is emerging as a key bottleneck. A new agentic artificial intelligence (AI) framework aims to address this challenge by automating aspects of variant curation and clinical interpretation while maintaining expert oversight.

Inocras (San Diego, CA, USA) is presenting the first public results from its multimodal agentic AI framework at the Association for Molecular Pathology (AMP) 2026 Annual Meeting & Expo, held November 10–14 in Seattle. The framework, which previews the company’s CancerVision workflow innovation, is designed to address a key scalability bottleneck in clinical cancer WGS analysis.

Image Credit: Adobe Stock

Rather than relying primarily on genomic annotations, the framework integrates multiple sources of evidence traditionally evaluated by expert curators. It conducts structured reviews of candidate driver variants using genomic annotations and sequencing metrics. When additional evidence is needed, the AI agent can visually inspect sequencing read alignments and retrieve information from an internal knowledge base of previously curated cases.

In its latest evaluation, the framework achieved 99.5% agreement with expert curation among accepted calls, while deferring 10.2% of candidate variants for expert review. It also demonstrated 98.4% reproducibility across repeated analyses using identical inputs and settings. The framework is projected to reduce expert review time by more than half, with human experts retaining responsibility for reviewing and signing off on all clinical reports.

The AMP presentation is titled “A Multimodal Agentic AI Framework for Scalable Variant Curation and Clinical Interpretation of Cancer Whole-Genome Sequencing.” It will be presented as oral presentation I024 in the Platform Presentations of Selected Informatics Abstracts I session on November 12, 2026. The same work is listed as poster I024, abstract 2465932, in the Informatics category, with poster display dates of November 12–14 and a presentation on November 13, 2026.

“Whole-genome sequencing offers a comprehensive view of cancer genomes, but interpreting the volume of data can be challenging. Our AI framework aims to streamline this process, making WGS more practical for clinical use by reducing curation workload while maintaining expert oversight,” said Jehee Suh, CEO of Inocras.

“Variant curation is becoming an increasingly important rate-limiting step as WGS moves toward broader clinical adoption. By combining multimodal reasoning, historical expert knowledge, and continuous feedback, our AI framework will be able to automate routine decisions while directing expert attention to the cases that require it most,” said WonChul Lee, CIO and Head of the Innovation Lab at Inocras.

Related Links
Inocras